NVIDIA AI Day Seoul brings together LG, NAVER, SK Telecom and Coupang around AI agents under their own control
NVIDIA held AI Day in Seoul with LG AI Research, NAVER Cloud, SK Telecom and Coupang to show how companies are building AI agents with open source tools — and...

Key takeaways
- A NVIDIA realizou o AI Day em Seul com LG AI Research, NAVER Cloud, SK Telecom e Coupang.
- O evento destacou agentes de IA construídos com ferramentas open source sob controle próprio.
- Empresas coreanas buscam autonomia no desenvolvimento de agentes de IA, evitando dependência de APIs fechadas.
- Desenvolvedores no Brasil podem se inspirar no modelo coreano de agentes de IA com ferramentas abertas.
NVIDIA met in Seoul, on AI Day, names such as LG AI Research, NOVER Cloud, SK Telecom and Coupang to discuss a specific point: how a company builds AI agents that it itself controls. The event, reported in the official NVIDIA blog, served less as a product showcase and more as a method demonstration — with the Korean AI ecosystem as a case study.
What did NVIDIA show about AI agents on Seoul's AI Day?
The meeting brought together four distinct profiles of companies: a research laboratory in AI (LG AI Research), a cloud arm (NOVER Cloud), a telecommunications operator (SK Telecom) and a giant electronic commerce (Coupang). The choice is not accidental. These are sectors where AI agents need to operate on internal processes, sensitive data, and scaled service — not just answering questions in a chat box.
According to NVIDIA, the focus of the presentations was the use of its open source tools for the development of corporate agents. The central message: the company must own the agent, not the tenant of a closed supplier.
Why did AI agents under their own control dominate the debate?
Because the question is no longer whether the AI works and becomes the boss of it. In corporate environments, an agent performing tasks needs permissions, audit trails, acting limits and controlled access to internal systems. When this agent runs within the perimeter of the company itself, governance becomes easier to describe — and easier to prove for audit and compliance.
This is the same dilemma that the agent governance market has been trying to solve on the side of security tools, as in the case of SailPoint Agentic Fabric in AWS Marketplace. The difference is the approach: instead of controlling the agent on the outside, the proposal discussed in Seoul is to build it on an open basis, which the company can inspect and adapt.
NVIDIA open source tools for AI agents: what do they deliver?
The source describes the role of these tools as a corporate agent development infrastructure, openly made available so that teams can assemble, adjust and integrate their own flows. The practical point is to reduce dependency: those who build on open source can exchange components, run in their own environment and keep data history inside the home.
It's worth the record of what the ad doesn't bring. There are no adoption numbers, performance benchmarks, prices or tool availability schedules in the material disclosed. NVIDIA positions the event as an exchange of experience between customers and the developer community, not as a metric release.
AI agents in Brazil: what changes for developers?
It changes at least by the ad and more by the set of references. The four sectors represented in Seoul — research, cloud, telecom and retail — are exactly the ones that invest most in AI in Brazil. For a national team, the message is that corporate agents have ceased to be a laboratory project: they have become an architectural item, with audit requirements that directly talk to LGPD.
There is also the cost factor. Open tools tend to weigh less on the budget of small teams, which cannot negotiate large-volume contracts. And the Korean experience leaves a sequence lesson: cloud and telecom first, because they are the ones that already have the data organized; retail after, focusing on service and operation.
AI agents under their own control: what the event does not solve
The discussion of self-control does not eliminate hard work. Internally constructed agent continues to require data curation, continuous evaluation of responses and maintenance — costs that do not appear in stage performances. In addition, there was no sign, at the source, of local partnerships in Brazil or of regulatory adaptations to the national market. The Seoul AI Day is for now a portrait of the Korean ecosystem. Translation to other markets is still open.
Frequently Asked Questions about AI and NVIDIA agents
What was NVIDIA AI Day in Seoul?
It was an NVIDIA event in the South Korean capital that brought together companies such as LG AI Research, NAVER Cloud, SK Telecom and Coupang to discuss the construction of AI agents controlled by the companies themselves, with support from open source tools.
Which companies participated in the AI Day?
NVIDIA cites LG AI Research, NAVER Cloud, SK Telecom and Coupang, as well as local developer communities, representing the research sectors in AI, cloud, telecommunications and e-commerce.
What does it mean to have an AI agent under control?
It means that the company keeps the agent within its own infrastructure and governance, defining permissions, limits of action and access to data, rather than rely exclusively on a closed platform of third parties.
Source: blogs.nvidia.com

Written by
Lucas MontarroiosSou Lucas Montarroios e dediquei os últimos 15 anos à linha de frente de operações de telecom e data centers. Minha carreira sempre foi pautada por um foco implacável: transformar tecnologia e cenários críticos em oportunidades reais de negócio. No novidades.ia.br, trago essa visão executiva para o universo da IA. Especialista em produtos, mercado e ferramentas práticas de IA. Minha missão aqui é filtrar o ruído do mercado, analisando benchmarks, estratégias de grandes empresas e ferramentas práticas.


